The world’s largest tech company is in talks with major media outlets to strike a landmark deal on the use of news content for training in artificial intelligence technology.
OpenAI, Google, Microsoft and Adobe have met with news executives in recent months to discuss copyright issues involving their AI products, including text chatbots and image generators, according to people familiar with the talks. .
Publishers including News Corp, Axel Springer, The New York Times and The Guardian are each in talks with at least one tech company, the people said.
Stakeholders to the talks, which are still in their early stages, said the agreement would include subscription-style fees for content by media organizations to develop the technology behind chatbots such as OpenAI’s ChatGPT and Google’s Bard. It added that it could include paying for
The talks come as media groups express concerns about the threats posed to the industry by the rise of AI and the uncontracted use of content by OpenAI and Google. Some companies, such as Stability AI and OpenAI, have faced lawsuits from artists, photo agencies and programmers alleging breach of contract and copyright infringement.
Speaking at the INMA media conference in May, News Corp CEO Robert Thomson summed up the industry’s outrage:[media’s] Collective intellectual property is under threat and needs to be vocally demanded for compensation. ”
He added that AI is “designed to ensure that readers never visit journalism websites, and that it fatally undermines journalism.”
The deal will set the blueprint for how news organizations will deal with generative AI companies around the world.
“Copyright is an important issue for all publishers,” said the Financial Times, which also discusses the issue. “As a subscription business, we must defend our journalism values and business model. Engaging in constructive dialogue with our affiliates, as we do, is the best way to achieve that.”
Media industry executives want to avoid the early internet mistakes that made so many articles available online for free, ultimately undermining their business models. Big tech groups such as Google and Facebook then accessed that information and helped build multi-billion dollar online advertising businesses.
As generative AI grows in popularity, so does the concern of the news industry, given the technology’s ability to generate compelling, human-like text.
Google recently announced a generative search feature that returns information boxes written by AI on top of the traditional web link format. We have launched in the US and are preparing for a worldwide release.
There are also current discussions to find a pricing model for news content to be used as training data for AI models. One figure that has been debated among publishers is between $5 million and $20 million a year, according to industry executives.
Matthias Dopfner, CEO of Politico owner Axel Springer, who met with AI giants Google, Microsoft and Open AI, said that “quantitative” technologies such as radio stations and nightclubs similar to those developed by the music industry have been developed. Building a model would be the first choice, he said. And streaming services pay record labels every time a track is played. That would first require AI companies to disclose how their media content is used, which they currently do not do.
Depfner, the Berlin-based media company that also owns German tabloid Bild and newspaper Die Welt, said an annual contract with unlimited use of the media company’s content would be “the next best option.” rice field. Because this model is difficult for small areas. Or use your local press.
“We need an industry-wide solution,” says Dopfner. “We have to work together on this.”
Google has led negotiations with UK media outlets and has held talks with the Guardian and NewsUK. The Alphabet company has long-term partnerships with many media organizations to use data from articles and other content to optimize their appearance in search engines. The company used the data to train language models at scale, according to two people familiar with the deal.
“Google put the licensing deal on the table,” said an executive at a newspaper group. “They have accepted the principle that they need to be paid . We acknowledge that there is, and that is the first step.”
Google will not comment on financial discussions. However, the search firm is in “continuous dialogue” with news outlets large and small in the US, UK and Europe, and is using AI on “public information” that may already include paywalled websites. Said he was training.
Just as it allows websites to opt out of using their content, the Silicon Valley giant gives publishers more “choice” about whether their content will be part of the AI’s training datasets. And added another option that I was looking at how to give “control”. search.
Since launching ChatGPT in November, OpenAI head Sam Altman has met with News Corp and The New York Times, according to people familiar with the discussions. The company confirmed that it has held discussions with publishers and publishing associations around the world on how it could work together.
Publishing industry leaders say it is very difficult to develop financial models to use news content to train AI. Executives at a major US publishing company said the news industry is working retroactively because technology companies launched these products without consultation.
“There was no discussion, so now we have to try to get paid after the fact,” the official said. “The way they launched these products, the complete secrecy, the zero transparency, the fact that there was no prior communication, there are reasons to be pretty pessimistic.”
Media analyst Claire Enders said the negotiations were “very complicated at the moment”, and with each body taking its own approach, the possibility of a single commercial deal for the media group is unlikely. Low and potentially counterproductive, he added.
Enders added, “Chatbots will not be reliable tools if they are literally primarily trained in the misogyny and racist sewage that dominates the vast majority of open and accessible text.” added.
Tech companies building AI are keen to focus on AI’s utility in streamlining newsrooms and enhancing journalism and are willing to pay millions to maintain longstanding relationships with industry negotiators said.
Brad Smith, vice chairman of Microsoft, said this is “in the early stages of dialogue with media and publishers, part of which is to help everyone learn about how models are trained.” It’s just there,” he said.
“I think the bigger opportunity for us is to work with publishers first to figure out how to leverage AI to generate more revenue,” he added.
Adobe CEO Shantanu Narayan has met with Disney, Sky and the UK’s Daily Telegraph in the past few weeks to discuss how each will develop custom models to harness their image-generating AI. said they had talked about
Adobe’s models are trained using images from our own library of stock images and openly licensed public domain content with expired copyrights. Narayan said bespoke deals and pricing are up to the company, but customers can add their own content to the tool.
Axel Springer’s Dopfner is optimistic that consensus will be reached as media organizations and policy makers grasp the scale of the challenge sooner than during the last major wave of technological disruption. expressed.
AI companies “know and fear regulation is coming,” he said, adding that “it’s in the interest of all parties to come up with solutions for a healthy ecosystem.” added. Without incentives to create intellectual property, there is nothing to crawl. And artificial intelligence will become an artificial fool. ”
